A scam reimbursement claim assessment template records what a bank needs to decide whether to refund a customer who was tricked into sending money: the payments, how the scam worked, warnings shown, the customer's circumstances, recovery attempts, the rules that apply and the decision with reasons. An AI agent assembles the file. A person decides.
Why does this assessment need a template?
When a customer authorises a payment because a fraudster deceived them, the money left with their consent. Whether the bank refunds them turns on rules that vary by market and on facts that vary by case: what the customer was told, what they could reasonably have known, whether they were vulnerable. Without a fixed structure, similar claims get different answers.
The decision itself carries weight a template cannot remove. That is the argument of scam reimbursement needs a person: the agent can make the file complete and consistent, but the judgement about the customer's conduct and circumstances belongs with someone accountable.
When should the template be opened?
The moment a customer reports a payment they made because of a scam. Early steps are time-critical: contacting the receiving bank quickly is the best chance of recovering funds. The agent should open the case, record the customer's account and draft the receiving-bank notification for a person to send within minutes, not days. The full process is in how to handle push payment scam reimbursement claims with an AI agent.
How do you complete each field?
- Customer's account. Record it in their words, once, without leading questions. It is declared evidence, and it matters.
- Payments. From the ledger, signed. Include every payment in the scam, not just the one reported.
- Warnings. What the bank's systems showed at the time, with the exact wording. The agent pulls this from the payment record.
- Vulnerability. Anything that affected the customer's ability to protect themselves. Flag only; never assess.
- Rules applied. Where a mandatory reimbursement scheme applies in your market, name it and its tests in your internal policy reference. Otherwise, your own policy.
- Decision and reasons. Written by the decision maker in plain words the customer will understand.
The template looks like an assessment of the customer, and it can be mistaken for one of the credit assessments banks run, but it is not. A creditworthiness assessment or an affordability assessment asks whether a customer can repay; this asks whether the bank should bear a loss the customer suffered.
The scam reimbursement claim assessment template
| No. | Field | Completed by |
|---|---|---|
| 1 | Claim reference, date reported, channel | Agent |
| 2 | Customer's account of events, in their words | Agent records |
| 3 | Payments: date, amount, payee, account, method | Agent |
| 4 | Scam type: purchase, investment, impersonation, romance, other | Agent proposes; person confirms |
| 5 | How the customer was first contacted | Agent records |
| 6 | Warnings shown at the time of payment, exact wording | Agent |
| 7 | Signs of vulnerability | Agent flags; person assesses |
| 8 | Receiving bank notified: time, by whom, reference | Person sends; agent drafts |
| 9 | Funds recovered | Agent tracks |
| 10 | Rules applied and policy reference | Person |
| 11 | Exceptions considered under those rules, with evidence | Person |
| 12 | Decision: full, partial or no reimbursement | Person |
| 13 | Reasons, in plain words for the customer | Person; agent may help with wording |
| 14 | Decision maker and date | Person |
A filled example: purchase scam
A hypothetical customer paid 1,500 for a used car advertised online and never received it.
| No. | Entry |
|---|---|
| 2 | "I saw the car online, the seller sent photos and a logbook picture, I paid the deposit and then the full amount. Then the number stopped working." |
| 3 | Two payments: 300 and 1,200, two days apart, to the same new payee |
| 4 | Purchase scam (confirmed by fraud analyst) |
| 6 | Generic new-payee warning shown on first payment; none on second |
| 7 | None flagged |
| 8 | Receiving bank notified 40 minutes after the report by fraud analyst |
| 9 | 200 recovered |
| 11 | Considered whether customer ignored a specific warning; warning shown was generic and did not describe purchase scams |
| 12 | Full reimbursement of the 1,300 not recovered |
| 13 | "We have refunded you because the warning we showed did not describe this type of scam." |
The agent assembled items 1 to 6 and 9 while the customer was still in the chat. The analyst made every judgement. Item 6, pulled directly from the payment record, decided the outcome.
Where Pryme fits
In Pryme, a claims agent for scam reimbursement holds depth 2 at most: it gathers, drafts and tracks, and the decision is held for a named analyst. The customer's account arrives stamped as declared. Before deploying, run the AI agent risk assessment template on the capability. The wider rules are in the guide to AI agents in banking.
Questions people ask
Can the agent tell the customer they will be refunded?
No. It can explain the process, what happens next and when they will hear. A promise before the decision would either bind the bank or mislead the customer.
Should the agent ask the customer probing questions?
It should ask the questions in the template, neutrally. Questions that test the customer's story should come from a person, because tone and judgement matter.
What about payments sent through several accounts?
Record every payment in field 3, including those through the customer's other accounts, because the scam is assessed as a whole.